On Tue, 21 Jun 2005, William Harold Newman wrote:
> and that's all she wrote, despite spending more than 10 CPU minutes.
> The program doesn't answer Ctrl-C at that point, so while probably
> there's some way to get gdb to tell me where the problem is, I don't
> happen to know it offhand (and haven't yet tried to educate myself).
> I've bisected the problem down to a change between .57 (which works)
> and .59 (which hangs the same way as most current). I should try .58
> too, but until I do, it seems like a reasonable bet that .58 is OK and
> the problem arrives with .59, because .59 touches so many *.[ch] files.
.59 indeed seems like the likely culprit.
However, this reminds me of a report in #lisp (which I hope we still hear
about from the horse's mouth to get the details):
Apparently clisp & gcc 3.3 on x86 Redhat (I've no more details) failed to
build 0.9.1.xx, dying with a gc_assert in purify.c from an invalid tag.
The same clisp and gcc apparently managed 0.9.0, which then built the
0.9.1.xx successfully with the same gcc.
...so possibly we have a leak. Does this happen with 0.9.0 as build host?
-- Nikodemus Schemer: "Buddha is small, clean, and serious."
Lispnik: "Buddha is big, has hairy armpits, and laughs."